MICRO- TO MACROSCALE MODELING OF DRINKING WATER TREATMENT AND DISTRIBUTION
Models are valuable tools that can provide quantitative insight into a process or system, but must to be kept simple to remain meaningful. In this dissertation, three separate drinking water research projects were completed using numerical modeling as a tool to describe and understand systems. Each project aimed to best exploit models or the modeling process while keeping the modeled descriptions and interpretations as simple as possible. These projects illustrate that modeling is a useful technique for gaining understanding of system structure and process behaviors, may guide experiments to verify hypotheses, and provides valuable data for assisting decision making. In the first project, five equations to calculate zeta potential for electrophoretic mobility measurements were compared to determine if complicated models return the same values determined by simple models. Results obtained by comparing models through analysis of electrophoretic mobility measurement uncertainty indicated that for 60% of the organisms studied, the Helmholtz-Smoluchowski or Henry equations produced zeta potential values that were not statistically different from equations which account for electrical double-layer distortion. The goal of the second project was to determine the process variables that are considered most important to model the removal of Cryptosporidium oocysts through the conventional drinking water treatment. Stepwise model analysis resulted in R2 values of 0.47 to 0.89. Although the models did not have high predictive values, stepwise analysis indicated the variables with the highest potential to predict Log Cryptosporidium removals. Results showed that observational variables like the portion of filter run and optimality of coagulant dose were important variables in the removal of Cryptosporidium regardless of the coagulant dose. Cryptosporidium removal appeared to be related to metal ion concentration below 0.4 mg/L (as ion), while above this value, metal ion concentration was not as important because other factors appear to have a stronger affect on oocyst removal. This study suggested it may be possible for drinking water treatment facilities adding less than 0.4 mg/L metal ion (e.g., direct filtration plants) to observe higher Cryptosporidium removals by increasing the dose of coagulant applied. In the third project, a distribution system model was developed to mimic the way it would be controlled by an operator. The model was then used to examine the causes of water aging in the system as well as operational scenarios that the utility could employ to improve the quality of water delivered to customers. This project provided an example of the techniques needed to use EPANET MSX for water age analysis and to examine the impact of individual tanks and their operation on system water age. Novel approaches to analyze data from hydraulic model simulations with respect to system water quality were used. For most of the water in the system, the age was dominantly controlled by the pipes (which contained 67% of total system volume) and only moderately influenced by the storage tanks (33% of total system volume). However, storage tank operation did show significant effects on the maximum water as well as the age of the water for the 3-5% of consumers receiving the oldest water.